Answers based on public websites
Import the service and procedure pages published by your municipality.
Portal Chat
For local government | Embeddable website AI chat
Find services
and procedures
by asking.
<script src=".../widget.js" async></script>FOR GOVERNMENT
Turn public municipal information
into guidance residents can use.
Guide residents through questions
Find the information residents need on public pages, based on their questions.
Check the sources
Show answers alongside source pages so users can return to the original information immediately.
INTRODUCTION FLOW
Import public websites.
Connect them to answers.
- 01Register target pages
Choose public websites covering the services and procedures you want to explain
- 02Import information
Retrieve public pages and organize them for search
- 03Install the chat
Add one script tag to your existing website
- 04Review answers
Check answers and source pages before launch
VISITOR EXPERIENCE
An easier experience,
with six practical features.
Suggested questions
Configure up to six questions per site to help users get started.
Streaming answers
Answers appear progressively through SSE, reducing the sense of waiting.
Reference pages
Source-page titles and URLs are displayed separately from the answer.
Follow-up conversations
Ask follow-up questions using the latest 20 history entries for context.
Switch topics
Reset the history to start another question with a fresh search context.
Readable answers
Display Markdown, tables, code, and links, with support for smartphones.
KNOWLEDGE & ANSWER PIPELINE
Import public pages.
Connect them to search and answers.
How public pages become knowledge
- 01Register target URLs
Set a name, URL, crawl depth, and maximum page count
- 02Crawl
Respect robots.txt, sitemaps, and Crawl-delay
- 03Extract content
Retrieve and store titles, body text, and links
- 04Make it searchable
Organize the captured information so AI can find it
- 05Store search data
Save the organized information ready for search
The crawler is designed for large sites, with progress stored in a database-backed frontier so interrupted crawls can resume.
How questions become answers with sources
- 01Question and history
Combine questions in the conversation to form a search query
- 02Find relevant information
Search the imported pages for information related to the question
- 03Generate an answer
Pass the retrieved context to an LLM to generate a response
- 04Answer and sources
Return the SSE answer and source pages separately
If retrieval does not work well, the chat continues using available information or switches to a response that takes safety into account.
ADMINISTRATION & OPERATIONS
From adding sites to improving operations,
manage it in one console.
Register and edit sites
Manage display names, URLs, crawl depth, and maximum page counts.
Monitor crawling and learning
See running, completed, and failed jobs along with error details in one list.
Review captured pages
Track titles, URLs, depth, HTTP status, and retrieval timestamps.
Question management
Review questions from both the widget and registered users, with details by question, source, user, and date.
Conversation history
Save conversation lists and message details to revisit multi-turn exchanges later.
User management
Add or remove administrators and regular users, change roles, and manage usage status.
IMPLEMENTATION
Get started
in four steps.
- 01Register your site
Set a name, URL, crawl depth, and maximum page count
- 02Crawl and learn
Retrieve public pages and organize them for AI search
- 03Add one line
Add the widget.js script tag to existing HTML
- 04Review and publish
Check CORS origins, answers, and sources before launch
CONTACT